The short answer
Really Good GEO groups the possible reasons into six diagnostic categories. These are hypotheses to investigate, not verified causes for any specific omission: retrievability, entity classification, question coverage, evidence support, competing sources, and test design.
1. Hypothesis: the page is hard to access
Blocking, script-only content, authentication, errors, or weak internal linking can prevent a useful page from becoming a candidate source.
2. Hypothesis: the company is hard to classify
Atmospheric language may never state the category, audience, geography, product type, or relationship between the company and its offer.
3. Hypothesis: the page does not answer the question
A feature list is not automatically an answer to “Which product is suitable for this operating constraint?”
4. Hypothesis: the claim lacks support
Claims without a method, source, qualification, specification, or first-hand example are harder for a reader — and harder for a model synthesizing a citation — to defend. This is a Really Good GEO framework principle, not a measured engine rule.
5. Hypothesis: better sources exist
Competing sources — competitors, directories, publishers, authoritative reference works — may resolve the question more completely, and models synthesizing an answer are likely to draw from them instead. This is a plausibility argument, not a measurement.
6. Hypothesis: the test is misleading
One run is noisy. A branded question measures known-entity retrieval, not whether an unknown buyer would discover the company.